Mir-19B Enhances Osteogenic Differentiation of Mesenchymal Stem

Home , Osteocalcin

Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 https://doi.org/10.1038/s12276-021-00631-w Experimental & Molecular Medicine

ARTICLE Open Access miR-19b enhances osteogenic differentiation of mesenchymal stem cells and promotes fracture healing through the WWP1/Smurf2-mediated KLF5/β-catenin signaling pathway Yan Huang1,2, Yongqiang Xu 1, Siyin Feng1,PanHe1, Bing Sheng1 and Jiangdong Ni2

Abstract Bone marrow mesenchymal stem cell (BMSC)-derived exosomes have been found to enhance fracture healing. In addition, microRNAs contributing to the healing of various bone fractures have attracted widespread attention in recent years, but knowledge of the mechanisms by which they act is still very limited. In this study, we clarified the function of altered microRNA-19b (miR-19b) expression in BMSCs in fracture healing. We modulated miR-19b expression via mimics/inhibitors in BMSCs and via agomirs in mice to explore the effects of these changes on osteogenic factors, bone cell mineralization and the healing status of modeled fractures. Through gain- and loss-of function assays, the binding affinity between miR-19b and WWP1/Smurf2 was identified and characterized to explain the underlying mechanism involving the KLF5/β-catenin signaling pathway. miR-19b promoted the differentiation of human BMSCs into osteoblasts by targeting WWP1 and Smurf2. Overexpression of WWP1 or Smurf2 degraded the target protein KLF5 in BMSCs through ubiquitination to inhibit fracture healing. KLF5 knockdown delayed fracture β

1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; healing by modulating the Wnt/ -catenin signaling pathway. Furthermore, miR-19b enhanced fracture healing via the KLF5/β-catenin signaling pathway by targeting WWP1 or Smurf2. Moreover, miR-19b was found to be enriched in BMSC-derived exosomes, and treatment with exosomes promoted fracture healing in vivo. Collectively, these results indicate that mesenchymal stem cell-derived exosomal miR-19b represses the expression of WWP1 or Smurf2 and elevates KLF5 expression through the Wnt/β-catenin signaling pathway, thereby facilitating fracture healing.

Introduction and biomineralization3. Fracture sites are commonly Bone fractures are the most frequently occurring type of subjected to increasing oxidative stress after injuries, large-organ, traumatic damage in humans1. The delayed which impairs osteoblast function and impedes the pro- healing process of bone fractures is a serious clinical and cess of fracture healing and remodeling4. The repair of economic problem for both patients and health services2. fracture wounds is critically influenced by mechanical The healing process of bone fractures is widely accepted loading as well as the geometric configuration of the as a complicated physiological course of events driven by fractured fragments5. The recent available evidence on the early inflammatory reactions and is accompanied by var- molecular mechanism indicates that bone fracture healing ious biological activities, such as osteogenic differentiation involves the complex coordination of various cell types, proteins and genes to restore structural integrity6. A limited amount of data have documented the paracrine Correspondence: Yongqiang Xu ([email protected]) mechanisms associated with the orchestration of 1 ’ Department of Orthopaedics, Hunan Provincial People s Hospital, Changsha, mesenchymal stem cell (MSC) transplantation in bone China 2Department of Orthopaedics, The Second Xiangya Hospital of Central South fractures, and exosomes are pivotal elements of the University, Changsha, China

© The Author(s) 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a linktotheCreativeCommons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 974

paracrine effects7. MSC-derived exosome-mediated delivery exosomes. BMSCs were cultured in exosome-free med- of microRNAs (miRNAs), such as miR-126, has shown ium containing 10% FBS, and BMSCs at passages 4-6 were potential in accelerating fracture healing8.Ofinterest, selected for exosome collection. A total of 2 × 106 MSCs upregulation of miR-19b could cause a significant increase were cultured in 100 mm culture dishes under normal/ in the transcription and translation of osteogenic factor hypoxic conditions for 72 h, and 10 ml of the supernatant genes during the process of osteogenic differentiation9.We was taken. Exosomes were then isolated from the super- used a bioinformatic prediction program to predict the natant after centrifugation16. Then, exosomes were potential target genes of miR-19b, and WWP1 and Smurf2 resuspended in PBS. were predicted as the candidate target genes of miR-19b. For identification of exosomes, samples were evaluated Notably, the PY motif-containing transcription factor under a JEM-2100 transmission electron microscope Kruppel like factor 5 (KLF5) is frequently degraded through (TEM; JEOL, Tokyo, Japan). Images were acquired using ubiquitination by E3 ubiquitin ligases such as WWP110.In the PARTICLEMEIRIX system. Nanoparticle tracking addition, prior evidence has suggested that SMAD ubiqui- analysis (NTA) was performed using the NanoSight tination regulatory factor 2 (Smurf2), an E3 ubiquitin ligase, NS300 system (Malvern Instruments, Malvern, UK). The is an interacting protein of KLF5 and diminishes the protein Brownian motion of exosomes in PBS was recorded and stability of KLF511. Furthermore, KLF5 has been revealed to tracked, and the size distribution was analyzed using the activate β-catenin signaling in breast cancer12.Inturn, Stokes-Einstein equation. The exosome characteristics Wnt/β-catenin signaling pathway activation has been pro- were identified by detecting the expression of the posed to enhance the production of osteogenic factors and exosome-specific surface markers with rabbit anti-CD63 thus expedite tibial fracture healing13.Inthisstudy,we (1:2000, ab216130, Abcam, UK), rabbit anti-TSG101 sought to explain the regulatory mechanism of exosomal (1:10,000, ab125011, Abcam), rabbit anti-CD81 (1:10,000, miR-19b derived from bone marrow mesenchymal stem ab109201, Abcam) and rabbit anti-Calnexin (1:100,000, cells (BMSCs) in the process of fracture healing, which may ab92573, Abcam) antibodies by Western blot analysis. involve the WWP1/Smurf2/KLF5/β-catenin axis. Experimental protocols Materials and methods When the cell density was 90% and the cells were in in Sample collection logarithmic growth phase, trypsin digestion and pipetting The bone marrow of healthy volunteers at The Second were performed to prepare a 2.5 × 104 cell/ml cell sus- Xiangya Hospital of Central South University was col- pension, which was added to a 6-well plate at 2 ml per lected, and informed consent was obtained from the par- well. When the cells were in logarithmic growth phase ticipants. This study was approved by the Ethics and were 30% confluent, the corresponding lentiviruses Committee of The Second Xiangya Hospital of Central (2 × 106 TU) and 5 μg of polybrene were added to 1 ml of South University. After surgery, portions of the fresh tissue serum/antibacterial drug-free medium. Forty-eight hours samples were stored in liquid nitrogen within 30 min and after transduction, 1 μg/ml puromycin was added to each were then transferred to a −80 °C freezer for further use. well to select stably transduced cells. After successful lentiviral infection, cells at a confluence of 70% to 80% Isolation and culture of human BMSCs were subjected to transient transduction of the miR-19b Human BMSCs were isolated from healthy volunteers mimic/inhibitor, WWP1 overexpression plasmid (oe- and amplified according to previously reported meth- WWP1), Smurf2 overexpression plasmid (oe-Smurf2), ods14. BMSCs below passage 4 were used in the following KLF5 overexpression plasmid (oe-KLF5), β-catenin over- experiments. Maintenance medium comprising dulbec- expression plasmid (oe-β-catenin), shRNA targeting KLF5 co’s modified Eagle’s medium (DMEM) (Invitrogen), 10% (sh-KLF5) or the corresponding negative controls (NCs) FBS and 100 U/ml penicillin/streptomycin (Invitrogen) according to the instructions for LipofectamineTM 2000 was employed to incubate BMSCs at 37 °C in 5% CO2. (Invitrogen). Cells were collected 48 h after transduction, The osteogenic differentiation induction medium was and the transduction efficiency was determined for sub- composed of 10% FBS, 50 mg/ml L-ascorbic acid, 10 mM sequent experiments. glycerophosphate, and 100 nM dexamethasone and anti- biotics (complete alpha-MEM) (Sigma, St. Louis, MO). Western blot analysis Detection was performed on the 7th day after osteogenic The transduced cells and fracture healing tissues har- differentiation was induced15. vested at 2 and 4 weeks after fracture were lysed with RIPA lysis and extraction buffer (Thermo Fisher Scien- Extraction and identification of exosomes from BMSCs tific). Protease inhibitors and phosphatase inhibitors were Alpha-MEM (Gibco, Green Island) containing 20% FBS added for lysis for 30 min. The protein concentration was (Gibco) was centrifuged for 18 h at 200,000 g to deplete determined using a Bio-Rad protein assay kit (Bio-Rad).

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 975

Extracts were resolved by SDS-PAGE and electroblotted The wild-type KLF5 and its mutant DNA template onto nitrocellulose membranes (Bio-Rad). The mem- translated in vitro were used by PCR using the forward branes were probed with primary antibodies against rabbit primer: 5′-GGATCCTAATACGACTCACTATAGGAAC WWP1 (ab104440), Smurf2 (ab94483), KLF5 (ab137676), AGACCACCATGGCTACAAGGGTGCTGAG-3′ (the β-catenin (ab32572), Runx2 (ab23981), Collagen I T7 promoter sequence is underlined; the start codon (ab34710), ALP (ab83259), and β-actin (1:1000, ab5694) ATG is in bold font) and reverse primer: 5′-TTTCT overnight at 4 °C. Next, the membranes were incubated AGACTACTTGTCATCGT CGTCCTTGTAATCGTTC with HRP-conjugated goat anti-rabbit IgG secondary TGGTGCCTCTTCATAT-3′ (the stop codon is in bold antibodies (ab6721). The antibodies were purchased from font). RNA transcription and protein translation were Abcam, Cambridge, UK. The blots were developed with carried out in vitro with Quick Coupled Transcription/ ECL reagent (BB-3501, Amersham, UK). Images were Translation Systems (Promega) using [35S]-methionine acquired with a Bio-Rad image analysis system (BIO- (Amersham Biosciences). RAD) and quantified with Quantity One v4.6.2 software. The purified GST fusion proteins (GST, GST-WWP1, Total cell protein and cytoplasmic protein levels were or GST-Smurf2) were pulled down in GST binding buffer internally normalized to β-actin levels to determine rela- and conjugated to 50 µl of 50% glutathione-Sepharose 4B tive protein levels. suspension beads. After incubation at 4 °C for 1 h, the cells were incubated with GST pulldown binding buffer. Ubiquitination assay Then, 10 µl of 35S-labeled in vitro translated protein For the ubiquitination assay, BMSCs were treated with (KLF5) was added and incubated at 4 °C for 2 h with the proteasome inhibitor MG132 (20 μM, ab141003, shaking. Bound proteins were eluted. GST-tagged pro- Abcam) for 4 h. Via incubation of the whole-cell lysate teins were visualized with Coomassie blue staining, and (500 μg protein/sample) with UbiCapture-Q Matrix 35S-labeled proteins were assessed by autoradiography. (Biomol) overnight under gentle stirring at 4 °C, all ubiquitinated proteins were pulled down. After three washes, In vivo protein ubiquitination conjugation assay the captured proteins were eluted with 2× SDS-PAGE The in vitro ubiquitination assay was carried out by loading buffer. Cycloheximide was added according to the using a ubiquitin-protein conjugation kit (Boston Bio- kit’s instruction manual (Sigma, St. Louis, Mo) at a con- chem, Cambridge, MA). A rabbit anti-reticulocyte centration of 100 ng/ml. Western blot analyses were lysate-translated, 35S-labeled KLF5 antibody was used adopted to examine the total KLF5 protein level at 0, 6, to observe the existence of GST-WWP1 or GST-Smurf2. 12, and 24 h. The samples were subjected to 10% SDS-PAGE for autoradiography. RNA immunoprecipitation (IP)-Western blot analysis Coimmunoprecipitation was performed using anti-myc Alizarin red staining (ARS) and alkaline phosphatase (ALP) and anti-FLAG antibodies, as well as an anti-FLAG staining monoclonal antibody (F1804, Sigma), anti-myc mono- After osteogenic differentiation induction for 2 weeks, clonal antibody (M5546, Sigma), and IgG goat anti-rabbit cells were washed twice with PBS and fixed with 4% par- polyclonal antibody (ab20272, 1:5000, Abcam, UK). Then, aformaldehyde. Cells were stained with 2% ARS (Sigma, transduced BMSCs were lysed on ice for 5 min. Then, the pH 4.2) to visualize matrix calcium deposition and were cell lysate was sonicated 4 times on ice (5 s each) and then subjected to ALP staining using nitroblue tetrazolium centrifuged at 4 °C for 10 min to harvest the supernatant (NBT)/5-bromo-4-chloro-3-indolyl-phosphate (Sigma). (200 µl), which was incubated with the primary anti-myc antibody (2 µl) at 4 °C overnight. Next, the solution was Luciferase activity assay incubated with 50% Protein A-agarose was at 4 °C for 2 h. The full-length WWP1 3′UTR sequence (WT) and The beads were washed five times with 500 µl of 1 × cell Smurf2 3’UTR sequence (WT) were obtained from the lysate. The protein was suspended in 50 µl of SDS sample NCBI database and inserted downstream of the firefly buffer and analyzed by Western blotting. luciferase gene driven by the SV40 promoter in the pEZX- MT01 vector to generate the recombinant vectors PEZX- GST pull-down assay MT01-3′UTR WWP1-Mut and PEZX-MT01-3′UTR The GST fusion protein was purified from bacterial Smurf2-Mut. PEZX-MT01-3′UTR WWP1-Wt or PEZX- DNA fragments. WWP1 and Smurf2 were amplified and MT01-3′UTR WWP1-Mut was cotransfected with the cloned into the pGEX-6p-1 GST fusion protein vector miR-19b mimic NC, miR-19b mimic, anti-miR-19b mimic (Amersham Biosciences). The purified GST fusion protein NC or anti-miR-19b into 293 T cells. Cells were cultured was eluted into 10 mm reduced glutathione. The protein for 48 h using Lipofectamine 2000 (Life Technologies). concentration was determined by the Bradford method. Luciferase activity was quantified using an EnVision 2102

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 976

Multilabel Plate Reader (Perkin Elmer Inc., Waltham, embedded in paraffin using the Leica EG embedding MA), with Renilla luciferase as the internal reference. system (Leica Microsystems, Wetzlar, Germany). The sections were stained with hematoxylin, eosin and Alcian Fracture model establishment and experimental grouping blue (Sigma). Histomorphological analysis was performed All animal experiments were performed in accordance using Zen2012 (Zeiss, Oberkochen, Germany), with eva- with the principles and procedures approved by the luation of the mineralized area. Laboratory Animal Ethics Committee of The Second Xiangya Hospital of Central South University. Transverse Quantification of gene expression femoral shaft fractures were generated in C57 male mice Total RNA was extracted using an RNeasy Mini Kit (8 weeks old) using a C-shaped instrument with a three- (Qiagen, Valencia, CA), and mRNA was reverse tran- point bend. The right knee was exposed by using the lateral scribed to cDNA using a reverse transcription kit metatarsal approach, and a medial metatarsal dislocation (RR047A, Takara, Japan). For miRNA detection, a miRNA was generated. The femoral intercondylar sulcus was First Strand cDNA Synthesis (Tailing Reaction) kit exposed at the knee joint by complete flexion, and a 0.5 mm (B532451-0020, Sangon, Shanghai, China) was used for diameter drill hole was made in the center of the inter- reverse transcription to obtain cDNA. Samples were loa- condylar sulcus. After minimal lateral exposure, a saw was ded using a SYBR® Premix Ex TaqTM II (Perfect Real used to cut a thin (3 mm deep) line along the central axis to Time) kit (DRR081, Takara, Japan) and were analyzed by weaken the bones. The right femur of each animal was qRT-PCR in a real-time quantitative PCR instrument (ABI fractured by a three-point bending technique17,18.The 7500, ABI, Foster City, CA). Three replicate wells were muscle fascia and skin were closed. After successful model established per sample. The miRNA universal negative establishment, the mice were randomly divided into 8 primer and the U6 internal reference upstream primer groups with 10 mice in each group. Except for those in the were provided in the miRNA First Strand cDNA Synthesis control group (sham operation), all mice were injected with (Tailing Reaction) kit. The other primer sequences for lentivirus-based transduction agents (25 mg/100 ml) carry- GAPDH and U6 as housekeeping genes are shown in ing the miR-19b agomir, oe-WWP1, oe-Smurf2, sh-KLF5 or Table 1 and were designed and provided by Sangon Bio- sh-β-catenin alone or in combination at the fracture site tech (Shanghai, China). The relative expression levels of ΔΔ after fracture. Photographing and identification of harvested the products were calculated using the formula 2- Ct. tissue samples were performed at different time points. Statistical analysis Injection of exosomes and grouping The data are expressed as the mean ± standard devia- After establishment of the fracture model using the tion values calculated by SPSS 21.0 statistical software abovementioned methods, the successfully established (IBM Corp, Armonk, NY). For comparisons between two model mice were randomly divided into three groups with groups, an independent samples t-test was conducted. For 10 mice in each group to explore whether BMSC-derived multiple comparisons, one-way analysis of variance exosomes accelerate fracture healing in vivo. After isola- (ANOVA) followed by Tukey’s post hoc test was per- tion and identification of exosomes from MSCs, 200 µg of formed. Differences were considered statistically sig- exosomes was injected around the fracture area, and the nificant when p < 0.05. mice were grouped into the control, PBS, and exo groups. Results X-ray examination for evaluation of the wound area miR-19b promotes the differentiation of human BMSCs Radiographic imaging was performed at 0 weeks (the into osteoblasts by targeting WWP1 and Smurf2 day of fracture) and 4 weeks after fracture. A digital X-ray Bioinformatic analysis predicted that miR-19b targets imaging system (30 kV, 8 mAs) was adopted to for X-ray WWP1 and Smurf2 and indicated that the 3′UTRs of imaging of the right femur fractures in the mice. Mice WWP1 and Smurf2 are complementary to the miR-19b were observed for fracture healing and callus growth, and seed region (Fig. 1a). This targeting effect exists in both the wound area was then measured. humans and mice. The luciferase reporter assay results (Fig. 1b, c) illu- Hematoxylin-eosin (HE)/Alcian blue staining for evaluation strated that miR-19b bound to the 3′UTR of WT WWP1 of mineralized callus area and Smurf2. Relative to that in the WWP1 3′UTR-WT + Femora were collected at the 4th week after fracture. miR-NC group, the luciferase activity in the WWP1 3′ These samples were fixed in 4% paraformaldehyde at 4 °C UTR-WT + miR-19b mimic group was appreciably for 24 h and embedded in paraffin. The bone specimens reduced (p < 0.05), while there was no significant differ- were fixed at room temperature and decalcified for ence in the corresponding two Mut groups (p > 0.05). 2 weeks. The decalcified bone was then dehydrated and Relative to that in the Smurf2 3′UTR-WT + miR-NC

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 977

Table 1 Primer sequences for qRT-PCR.

Gene Human Mouse

WWP1 F: 5′-GTATGGATCCTGTACGGCAGCA-3′ F: 5′-ACCCAAGAGCCTCCTGTA-3′ R: 5′-GTTGTGGTCTCTCCCATGTGGT-3′ R: 5′-CTCTGTTCCCACCCTGAT-3′ Smurf2 F: 5′-GGCTCAATTCTTGGCTCTG-3′ F: 5′-CAGCACCTGCTGAAGACATT-3′ R: 5′-ACCGGGTGTTTACCTTCC-3′ R: 5′-GAACCACTTGACGACAT TGC-3′ KLF5 F: 5′-TCCACGCCACTAAGCATGTG-′ F: 5′-ACTGCCCTCGGAGGAGCTGG-3′ R: 5′-CGTAAATGACCGTCCTGGTCTT-3′ R: 5′-ATGCTCTGAAATTATCGGAACTG-3′ β-catenin F: 5′-AGTTGAGCACCTGTTTGCCTGA-3′ F: 5′-GCATCAATGGCTCAAGGACAAG-3′ R: 5′-ATGAGCAGCACTCGGACCTTC-3′ R: 5′-CCGGTCTCCAATTCCCAAGATA-3′ hsa-miR-19b 5′-GCAAATCCATGCAAAACTGA-3′

F forward, R reverse.

Fig. 1 miR-19b promotes the differentiation of human BMSCs into osteoblasts by targeting WWP1 and Smurf2. a Bioinformatic analysis (TargetScan) predicted that the 3′UTRs of WWP1 and Smurf2 are complementary to the miR-19b seed region. b The wild-type WWP1 3′UTR (named WWP1 3’UTR-WT) and mutant WWP1 3′UTR (named WWP1 3′UTR-MUT; UUGCAC was mutated to CAUUGG) were constructed with the miR-19b binding sites, and luciferase activity was measured with a dual luciferase reporter assay. c The wild-type Smurf2 3′UTR (named Smurf2 3′UTR-WT) and mutant Smurf2 3′UTR (named Smurf2 3′UTR-MUT; UUUGCA was mutated to ACGUUU) were constructed with the miR-19b binding sites, and luciferase activity was measured with a dual luciferase reporter assay. d The expression level of miR-19b and mRNA expression of WWP1 and Smurf2 in human BMSCs was detected by qRT-PCR, *p < 0.05 versus the mimic-NC group, #p < 0.05 versus the inhibitor-NC group, &p < 0.05 versus the miR- 19b mimic + oe-NC group. e The protein levels of WWP1, Smurf2, Runx2, ALP, and Collagen I in human BMSCs were determined by Western blot analysis. f Quantitative analysis of Alizarin red staining and ALP staining was performed to determine the degree of mineralization in human BMSCs. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For comparisons between two groups, an independent samples t test was conducted. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. The experiment was repeated 3 times. group, the luciferase activity in the Smurf2 3′UTR-WT + Subsequent experiments were performed using human miR-19b mimic group was appreciably reduced (p < 0.05), BMSCs. In response to miR-19b mimic treatment, the and no signiﬁcant difference was observed in luciferase mRNA levels of WWP1 and Smurf2 were appreciably activity in the corresponding two Mut groups (p > 0.05). reduced but were elevated in the presence of the miR-19b

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 978

inhibitor. In addition, the mRNA level of WWP1 or KLF5 ubiquitination after overexpression of WWP1 or Smurf2 was increased in response to combined treatment Smurf2 was assessed by adding ubiquitin (Fig. 2f, g). with miR-19b mimic + oe-WWP1 or miR-19b mimic + WWP1 and Smurf2 were adopted to induce ubiquitina- oe-Smurf2, respectively, but the influence of the miR-19b tion of endogenous KLF5 protein in cells. After treatment mimic on the WWP1 or Smurf2 mRNA level was abol- with MG132 (20 μM, 4 h), ubiquitination of KLF5 by ished in the presence of the Smurf2 or WWP1 over- WWP1 or Smurf2 was observed during incubation with expression plasmid, respectively (Fig. 1d). Consistent myc-tagged WWP1 or Smurf2, respectively, via Western results were observed in the Western blot analysis; miR- blot analysis with an anti-KLF5 antibody. When Western 19b mimic treatment resulted in decreased protein levels blot analysis was performed using an anti-HA tag anti- of WWP1 and Smurf2, and treatment with the miR-19b body for labeling, the ubiquitination of KLF5 by myc- inhibitor increased these levels. Furthermore, the miR- tagged WWP1 or Smurf2 was more obvious. These results 19b mimic led to increases in the protein levels of Runx2, provide direct evidence that WWP1 and Smurf2 are the ALP, and Collagen I, and the miR-19b inhibitor caused E3 ubiquitin ligases of KLF5. reductions in these levels (Fig. 1e). The ARS staining and After co-transduction of oe-WWP1 or oe-Smurf2 and ALP staining results (Fig. 1f) revealed a reduced propor- oe-KLF5, the changes in the Runx2, ALP, and Collagen I tion of mineralization in response to the miR-19b inhi- protein levels were assessed by Western blot analysis bitor; in contrast, mineralization was promoted by the (Fig. 2h). Treatment with oe-WWP1 or oe-Smurf2 alone miR-19b mimic. Treatment with oe-WWP1 and oe- diminished the protein levels of Runx2, ALP and Col- Smurf2 suppressed the miR-19b mimic-induced increase lagen I. However, combined treatment with oe-KLF5 and in the proportion of mineralization. These data indicate oe-WWP1 or oe-Smurf2 led to higher protein levels of that miR-19b targets and inhibits WWP1 and Smurf2 Runx2, ALP and Collagen I than those generated in expression to enhance the differentiation of human response to treatment with oe-WWP1 or oe-Smurf2 BMSCs into osteoblasts. alone. The data suggested that WWP1 and Smurf2 degraded their target protein KLF5 through ubiquitina- WWP1 or Smurf2 degrades KLF5 by ubiquitination to tion to inhibit fracture healing. repress fracture healing Next, we sought to explain the mechanism of the E3 KLF5 knockdown delays fracture healing by modulating ligases WWP1 and Smurf2 on the ubiquitination of KLF5 the Wnt/β-catenin signaling pathway in BMSCs. The transduction efficiency in BMSCs was Evidence indicates that KLF5 is involved in the regula- evaluated after transduction of the WWP1 and Smurf2 tion of β-catenin in breast cancer12. Additionally, it has overexpression plasmids. The PCR results validated the been reported that the Wnt/β-catenin signaling pathway good transduction efficiency of oe-WWP1 and oe-Smurf2. is involved in fracture repair13. Furthermore, we clarified In addition, treatment with oe-WWP1 and oe-Smurf2 whether KLF5 plays a role in modulating the Wnt/ resulted in reduced KLF5 expression (Fig. 2a). Western blot β-catenin signaling pathway during fracture healing. We analyses were performed at 0, 6, 12, and 24 h to examine downregulated KLF5 and measured the mRNA levels of total KLF5 protein levels, as shown in Fig. 2b. After over- the Wnt/β-catenin signaling pathway-related factor expression of WWP1 or Smurf2, the degradation rate of β-catenin in cells by qPCR. As shown in Fig. 3a, sh-KLF5 the KLF5 protein in BMSCs increased with time. treatment reduced the mRNA levels of KLF5 and IP assays and GST pulldown assays were carried out to β-catenin, and oe-β-catenin appreciably increased the determine whether WWP1 or Smurf2 directly interacts level of β-catenin that was decreased by sh-KLF5. In with KLF5 at the protein level. Myc-tagged WWP1 and addition, consistent results were revealed by Western blot Smurf2 were expressed in cells to detect endogenous analysis, which verified that oe-β-catenin markedly ele- KLF5. As shown in Fig. 2c, myc-WWP1 and myc-Smurf2 vated the protein level of β-catenin that was decreased by were immunoprecipitated by the anti-KLF5 antibody but sh-KLF5. Furthermore, the decreases in the protein levels not by IgG as the control. These results indicate that of Runx2, ALP, and Collagen I induced by sh-KLF5 were WWP1 and Smurf2 interact with endogenous KLF5 in rescued by oe-β-catenin (Fig. 3b). These data indicate that BMSCs. A GST pulldown assay was carried out using the KLF5 knockdown delays fracture healing by modulating GST-WWP1 and GST-Smurf2 fusion proteins and the in the Wnt/β-catenin signaling pathway. vitro-translated KLF5 protein. The wild-type proteins (GST-WWP1 and GST-Smurf2) pulled down the KLF5 miR-19b enhances fracture healing via the KLF5/β-catenin protein, but the GST protein alone could not bind to the signaling pathway by targeting WWP1 or Smurf2 KLF5 protein under the same conditions (Fig. 2d, e). Primary human BMSCs were co-transduced with the The ubiquitination assay was performed under the IP miR-19b mimic and sh-KLF5 or sh-β-catenin, and experimental conditions with FLAG, and the effect on osteogenic differentiation was then induced for 7 days.

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 979

a b c d oe-NC Control-WWP1 GST-WWP1 131 kDa oe-WWP1 oe-WWP1 IP 35S-KLF5 50 kDa oe-Smurf2 Control-Smurf2 4 1.5 Myc-WWP1 105 kDa T 1 * oe-Smurf2 Input GS WP Myc-Smurf2 86 kDa W 3 T- 1.0 GS * KLF5 50 kDa e 2 GST-Smurf2 112 kDa IgG * KLF5 Input * 0.5 * 35S-KLF5 50 kDa 1 * t

Relative expression u * ST rf2 u * * * Inp G * * 0 0.0 T-Sm WWP1 Smurf2 KLF5 Relative protein expression of KLF5 S 0 h 6 h 12 h 24 h G MG132 ++ MG132 ++ f HA-Ub ++ g HA-Ub ++ h KLF5-FLAG ++ KLF5-FLAG + + Myc-WWP1 -+KDa Myc-Smurf2 -+KDa

oe-NC IP: FLAG oe-WWP1 Ubiquitinated IB: KLF5 -100 Ubiquitinated oe-WWP1 + oe-KLF5 IP: FLAG -100 KLF5 KLF5 IB: KLF5 -75 1.5 oe-Smurf2 oe-Smurf2 + oe-KLF5 -75

& # KLF5- 1.0 # -50 KLF5- -50 &

Ubiquitinated Ubiquitinated # & KLF5 KLF5 0.5 * -100 -100 * * * * IP: FLAG IP: FLAG * IB: HA IB: HA -75 Relative protein expression -75 0.0 Runx2 ALP Collagen I -50 -50 Myc- -105 Myc- -105 FLAG- -50 FLAG- -50 β-actin- -42 β-actin- -42

Fig. 2 Overexpression of WWP1 or Smurf2 degrades their target protein KLF5 in BMSCs through ubiquitination to inhibit fracture healing. a PCR was adopted to examine the mRNA levels of WWP1, Smurf2, and KLF5. *p < 0.05 versus the oe-NC group. b After cycloheximide treatment (100 ng/ml), the total protein level of KLF5 was measured by Western blot analysis at 0, 6, 12, and 24 h. c Myc was used to label WWP1 and Smurf2, and endogenous KLF5 was detected with input as the positive control and IgG as the negative control. d–e GST pulldown assays were performed using GST-WWP1 (d) or GST-Smurf2 (e) fusion proteins and in vitro-translated KLF5 protein to examine exogenous KLF5 expression. Input is the positive control, and GST is the negative control. f–g Ubiquitin was added to examine ubiquitination of KLF5 after overexpression of WWP1 (f)or Smurf2 (g). h Western blot analysis of the protein levels of Runx2, ALP, and Collagen I with β-actin as the internal reference protein. The relative protein expression levels of Runx2, ALP, and Collagen I. *p < 0.05 versus the oe-NC group, #p < 0.05 versus the oe-WWP1 + oe-KLF5 group, &p < 0.05 versus the oe-Smurf2 + oe-KLF5 group. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. The experiment was repeated 3 times.

The miR-19b mimic successfully induced miR-19b over- 19b mimic + sh-β-catenin diminished the β-catenin level expression, which in turn resulted in reductions in the but did not signiﬁcantly alter the level of KLF5. Western mRNA levels of WWP1 and Smurf2. However, combined blot analysis, as shown in Fig. 4b, c, veriﬁed that the miR- treatment with miR-19b mimic + sh-KLF5/sh-β-catenin 19b mimic decreased the levels of WWP1 and Smurf2, yet abolished the effects of the miR-19b mimic on the levels the addition of oe-WWP1 to the miR-19b mimic upre- of miR-19b, WWP1 and Smurf2 (Fig. 4a). gulated WWP1, and combined treatment with the miR- Primary human BMSCs were transfected with the miR- 19b mimic and oe-Smurf2 upregulated Smurf2 in BMSCs. 19b mimic, oe-WWP1, oe-Smurf2, or sh-β-catenin alone Additionally, the miR-19b mimic elevated the levels of or in combination. The results of qPCR after 7 days of Runx2, ALP, and Collagen I, while this miR-19b mimic- osteogenic differentiation are shown in Supplementary induced upregulation was appreciably reduced by sh-KLF5 Fig. S1a. It was shown that overexpressing miR-19b ele- or sh-β-catenin (Fig. 4d). In addition, the ARS staining and vated the mRNA expression levels of KLF5 and β-catenin ALP staining results (Fig. 4e) revealed that the proportion and that this miR-19b mimic-induced upregulation of of mineralization was increased by the miR-19b mimic and expression was appreciably reduced by oe-WWP1, oe- that this increase was reversed by oe-WWP1, oe-Smurf2, Smurf2, or sh-KLF5. Combined treatment with the miR- or sh-KLF5. These data indicate that miR-19b can enhance

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 980

a sh-NC + oe-NC b sh-NC + oe-NC sh-KLF5 + oe-NC sh-KLF5 + oe-NC 1.5 sh-KLF5 + oe-β-catenin 1.5 sh-KLF5 + oe-β-catenin #

# 1.0 1.0 #

# # 0.5 0.5 * * * * * * *

Relative expression * *

0.0 Relative protein expression 0.0 KLF5 β-catenin KLF5 β-catenin Runx2 ALP Collagen I

Fig. 3 KLF5 knockdown delays fracture healing by modulating the Wnt/β-catenin signaling pathway. Primary human BMSCs were transduced with lentiviral vectors, and osteogenic differentiation was induced for 7 days. a The mRNA levels of KLF5 and β-catenin were measured by qPCR. *p < 0.05 versus the sh-NC + oe-NC group, #p < 0.05 versus the sh-KLF5 + oe-NC group. b Western blot analysis of the protein levels of KLF5, β-catenin, Runx2, ALP, and Collagen I, with β-actin as the internal reference protein. *p < 0.05 versus the sh-NC + oe-NC group, #p < 0.05 versus the sh-KLF5 + oe-NC group. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. The experiment was repeated 3 times. fracture healing by targeting WWP1 or Smurf2 expression Quantification of the healing area showed that relative to and modulating the KLF5/β-catenin signaling pathway. the control group at week 4, the miR-19b agomir group had a larger healing area; in addition, the remodeling process miR-19b enhances fracture healing in mice through the was observed, indicating that bone repair was accelerated. WWP1/Smurf2/KLF5/β-catenin axis The healing area in the oe-WWP1 group and the oe- To further confirm the effect of miR-19b on fracture Smurf2 group at the 4th week were not appreciably dif- healing in vivo, a femoral fracture model was established in ferent from the healing area in the control group, indicating mice. X-ray and micro-computed tomography (μCT) ima- that bone repair was slow. Moreover, there was no sig- ging were adopted to assess the femoral fractures (Fig. 5a) nificant difference in fracture healing at the 4th week in the (Due to the large number of groups, the data in the Fig. 5 miR-19b agomir + oe-WWP1group,themiR-19bagomir only show the untreated group of the fracture model as the + oe-Smurf2 group, the miR-19b agomir + sh-KLF5 group, Control group; the other control groups are not shown). On or the miR-19b agomir + sh-β-catenin group relative to the the 7th day after injection, the calli were harvested from the control group. mice for evaluation of miR-19b expression to confirm the At the 4th week after fracture, fracture calli were dec- transduction efficiency (Fig. 5a). In the mice, treatment with alcified for tissue sectioning and was then stained with HE/ the miR-19b agomir resulted in diminished protein levels of Alcian blue for quantification of wound healing (Fig. 5d; WWP1 and Smurf2 and elevated protein levels of KLF5 and Supplementary Fig. S1c). At the 4th week, miR-19b agomir β-catenin. Reductions were observed in the protein levels of treatment had resulted in a higher mineralization rate, KLF5 and β-catenin in response to oe-WWP1 and oe- better fracture wound healing, and accelerated fracture Smurf2. In addition, combined treatment with the miR-19b healing. Treatment with oe-WWP1 and oe-Smurf2 caused agomir + oe-WWP1 or oe-Smurf2 led to increased levels of lower mineralization rates, slower wound healing and a WWP1 or Smurf2, respectively, and decreased levels of later onset of fracture healing. The results of combined KLF5 and β-catenin but did not affect the levels of Smurf2 treatment with miR-19b agomir + oe-WWP1, miR-19b or WWP1. Combined treatment with miR-19b agomir + agomir + oe-Smurf2, miR-19b agomir + sh-KLF5, and sh-KLF5 diminished the KLF5 and β-catenin levels but did miR-19b agomir + sh-β-catenin were not appreciably not affect the WWP1 and Smurf2 levels. Combined treat- different from those of the control treatment. Among ment with the miR-19b agomir + sh-β-catenin diminished these results, relative to the oe-WWP1 group, the miR-19b the β-catenin level but did not affect the level of KLF5, agomir + oe-WWP1 group had a higher mineralization WWP1 or Smurf2 (Fig. 5b). Together, these results indicate rate and better fracture wound healing. Relative to the oe- that miR-19b enhances KLF5/β-catenin expression by Smurf2 group, the miR-19b agomir + oe-Smurf2 group inhibiting WWP1/Smurf2 in vivo. had a higher mineralization rate and better fracture wound At 0 and 4 weeks after the fracture model was established, recovery. Relative to miR-19b agomir treatment, the miR- the healing area of internal bone formation during the 19b agomir + sh-KLF5 and miR-19b agomir + sh- healing process of the femoral fractures in mice was eval- β-catenin treatments resulted in a lower mineralization uated by X-ray imaging (Fig. 5c; Supplementary Fig. S1b). rate and slower fracture wound healing.

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 981

mimic-NC + oe-NC mimic-NC + oe-NC a bcmiR-19b mimic + oe-NC miR-19b mimic + oe-NC miR-19b mimic + oe-WWP1 miR-19b mimic + oe-WWP1

mimic-NC + sh-NC miR-19b mimic + oe-Smurf2 miR-19b mimic + oe-Smurf2 miR-19b mimic + sh-NC miR-19b mimic + sh-NC miR-19b mimic + sh-NC miR-19b mimic + sh-KLF5 miR-19b mimic + sh-KLF5 miR-19b mimic + sh-KLF5 miR-19b mimic + sh-β-catenin miR-19b mimic + sh-β-catenin miR-19b mimic + sh-β-catenin 8 2.0 1.0 7 * N.S. * 0.8 # 6 1.5 * * * 5 N.S. 0.6 4 # 2 1.0 0.4 N.S. N.S. * # * # & * 0.5 # * * 1 N.S. N.S. # & & 0.2 Relative expression Relative protein expression

0 Relative protein expression 0.0 0.0 miR-19b WWP1 Smurf2 WWP1 Smurf2 KLF5 β-catenin d e mimic-NC + oe-NC mimic-NC + oe-NC miR-19b mimic + oe-NC miR-19b mimic + oe-NC miR-19b mimic + oe-WWP1 miR-19b mimic + oe-WWP1 mimic-NC + sh-NC miR-19b mimic + oe-Smurf2 miR-19b mimic + oe-Smurf2 miR-19b mimic + sh-NC miR-19b mimic + sh-NC miR-19b mimic + sh-NC miR-19b mimic + sh-KLF5 miR-19b mimic + sh-KLF5 miR-19b mimic + sh-KLF5 miR-19b mimic + sh-β-catenin 2.0 miR-19b mimic + sh-β-catenin 0.8 miR-19b mimic + sh-β-catenin * 0.8 * * 1.5 * * 0.6 0.6 * * & # # # & 1.0 # 0.4 # 0.4 # & & # # # # 0.5 of Alizarin0.2 red 0.2 Relative ALP activity Semi-quantitative result Relative protein expression 0.0 0.0 0.0 Runx2 ALP Collagen I

Fig. 4 miR-19b enhances fracture healing via the KLF5/β-catenin signaling pathway by targeting WWP1 or Smurf2. Primary human BMSCs were transduced with lentiviral vectors followed by induction of osteogenic differentiation for 7 days. a The expression level of miR-19b and mRNA levels of WWP1 and Smurf2 were measured by qPCR. *p < 0.05 versus the mimic-NC + sh-NC group, NS p > 0.05 versus the miR-19b mimic + sh-NC group. b–c Western blot analysis of the protein levels of KLF5 (b), β-catenin (b), WWP1 (c), and Smurf2 (d), with β-actin as the internal reference protein. *p < 0.05 versus the mimic-NC + sh-NC group, #p < 0.05 versus the miR-19b mimic + oe-NC, &p < 0.05 versus the miR-19b mimic + sh-NC group, NS p > 0.05. d Western blot analysis of the protein levels of Runx2, ALP, and Collagen I, with β-actin as the internal reference protein. *p < 0.05 versus the mimic-NC + sh-NC group, #p < 0.05 versus the miR-19b mimic + sh-NC group. e Quantitative analysis of ARS staining and ALP staining was performed to determine the degree of cell mineralization. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. The experiment was repeated 3 times.

These data suggest that the miR-19b agomir accelerates (Supplementary Fig. S2a), and the number of exosomes fracture healing, that oe-WWP1 and oe-Smurf2 delay increased after induction. Dynamic light scattering (DLS) fracture healing, and that the miR-19b agomir + oe- analysis data suggested that the diameter of MSC-exos WWP1 and miR-19b agomir + oe-Smurf2 treatments can was mainly between 30-100 nm (Supplementary Fig. S2b). reverse the delay in healing induced by oe-WWP1 and oe- Western blot detection of exosome surface markers also Smurf2. miR-19b activates the KLF5/β-catenin signaling revealed that the expression levels of exosome markers pathway through inhibition of WWP1/Smurf2, accel- (CD9, CD81, TSG101, HSP70, ALIX and Flotillin-1) in erating fracture healing in mice. MSCs were increased (Supplementary Fig. S2c). We also detected the expression of miR-19b in BMSC-exos, and BMSC-exos accelerate fracture healing in vivo by the data suggested that the expression of miR-19b was delivering miR-19b higher than that in MSCs (Supplementary Fig. S2d). In this study, exosomes derived from BMSCs (BMSC- On the 7th day after animal model establishment, calli exos) were characterized. TEM observation showed that were harvested from mice to quantify the expression of BMSC-exos exhibited a cup-like or spherical morphology miR-19b. The miR-19b expression level in the exo group

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 982

a b 4 2.0 Control Control oe-WWP1 * * 1.5 3 miR-19b agomir miR-19b agomir + oe-WWP1 miR-19b agomir + oe-WWP1 oe-Smurf2 2 miR-19b agomir + oe-Smurf2 1.0 * * # miR-19b agomir + oe-Smurf2 miR-19b agomir + sh-KLF5 * # # # # miR-19b agomir # # # 1 0.5 N.S. # miR-19b agomir + sh-β-catenin N.S. * * miR-19b agomir + sh-KLF5 * * * * miR-19b agomir + sh-β-catenin 0 Relative protein expression 0.0 Relative expression of miR-19b WWP1 Smurf2 KLF5 β-catenin

Control miR-19b agomir miR-19b agomir + sh-KLF5 miR-19b agomir + sh-β-catenin c Control miR-19b agomir miR-19b agomir + sh-KLF5 miR-19b agomir + sh-β-catenin

0 W 8 * ) 2 6

# 4 #

2 Callus Area (mm

4 W Control miR-19b agomir miR-19b agomir + sh-KLF5 miR-19b agomir + sh-β-catenin

d Control miR-19b agomir miR-19b agomir + sh-KLF5 miR-19b agomir + sh-β-catenin 100 * 80 # # 60

25 μm 25 μm 25 μm 25 μm Wound healing rate (%) 0

Fig. 5 miR-19b promotes fracture healing in mice through the KLF5/β-catenin axis. Mice in the femoral fracture model were injected with transduction reagents after the operation. a On the 7th day after injection and in week 0 after fracture, calli were harvested from mice in each group, and the level of miR-19b was measured by PCR to confirm the transduction efficiency of the miR-19b agomir in mice. *p < 0.05 versus the control group. b In week 0 after fracture, calli were harvested from mice in each group, and the protein levels of WWP1, Smurf2, KLF5, and β-catenin were measured by Western blot analysis, with β-actin as the internal reference protein. *p < 0.05 versus the control group, #p < 0.05 versus the miR-19b agomir group, NS > 0.05. c Representative X-ray images acquired to monitor femoral healing and changes over time at weeks 0 and 4 after fracture. d At the 4th week after femoral fracture, callus sections were stained with HE/Alcian blue, and the histology of the fracture calli was observed. Blue indicates cartilage, and pink indicates mineralized callus area. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. n = 5 in the weeks 0 and 4 after fracture. was higher than that in the PBS and control groups (Fig. remodeling process was observed. At the 4th week after 6a). In addition, the protein expression levels of WWP1, fracture in mice, the fracture calli from the treated mice in Smurf2, KLF5 and β-catenin were measured in mice of each group were decalcified and stained with HE/Alcian each group (Fig. 6b). The data suggested that WWP1 and blue. As shown in Fig. 6d, at the 4th week, the exo group Smurf2 were downregulated and KLF5 and β-catenin had a higher mineralization proportion and better were upregulated in the exo group relative to the control recovery ability of fracture wounds than the control group. These data suggest that injection of BMSC-exos group. The above data suggested that BMSC-exos upre- can promote the expression of KLF5/β-catenin by inhi- gulated miR-19b expression and accelerated fracture biting WWP1/Smurf2. At week 4 after the establishment healing in mice. of the fracture model, the healing area of the inner bone formation during femoral fracture healing in mice was Discussion evaluated by X-ray imaging (Fig. 6c). Statistical analysis Bones have repair potential following fractures, and the suggested that the healing area was larger in the exo healing process is involved in bone repair and the group than in the control group at the 4th week, and the restoration of structural integrity19. Healing of bone

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 983

ab c 8 4 1.5 PBS # * # # Control * * ) # 2 6 3 exo 1.0 *

2 4

0.5 1 # 2 * # Callus Area (mm * Relative protein expression

Relative expression of0 miR-19b 0.0 0 PBS Control exo WWP1 Smurf2 KLF5 β-catenin PBS Control exo

# d 100 *

PBS Control exo 80

20 Wound healing rate (%) 0 PBS Control exo

Fig. 6 BMSC-exos promote fracture healing in mice. a PCR was used to detect the level of miR-19b. b Western blot analysis of the protein levels of WWP1, Smurf2, KLF5, and β-catenin in the calli of each group at week 0 after fracture, with β-actin as the internal reference protein. *p < 0.05 compared with the PBS group; #p < 0.05 compared with the control group; NS p > 0.05. c Quantitative analysis of the callus area during femoral fracture healing at the 4th week after fracture. d HE/Alcian blue staining results for evaluating the mineralized callus area. The experimental results are measurement data and are expressed as the mean ± standard deviation values. For multiple comparisons, one-way ANOVA followed by Tukey’s post hoc test was performed. n = 5 in the weeks 0 and 4 after fracture. fractures restores the preinjury structure and function of mineralization and healing processes in mice. Consistent the damaged skeleton, but nearly 10% of fractures cannot results were reported in a prior study, which emphasized be repaired normally1. BMSCs possess multilineage dif- that osteogenic differentiation was induced by ectopic ferentiation capabilities, including osteogenesis and miR-19b-3p expression, corresponding to elevated chondrogenesis20, which enable them to function as osteocalcin, osteopontin, and Runx2 levels9. mediators to facilitate fracture repair21. The current study Furthermore, the targeting relationship between miR- explored the role of exosomal miR-19b derived from 19b and WWP1/Smurf2 was determined by an initial BMSCs in fracture healing. bioinformatic prediction followed by a confirmatory dual- The experimental data revealed that miR-19b promoted luciferase reporter assay. miRNAs have the capacity to the differentiation of human BMSCs into osteoblasts via orchestrate gene expression by binding to complementary targeting WWP1 and Smurf2. Notable evidence was sites in mRNAs and reducing the stability and translation obtained demonstrating that miRNAs play a pivotal role of the target mRNAs26. Ubiquitin E3 ligase-dependent in bone homeostasis by orchestrating the expression of protein degradation facilitates proteasomal degradation of key genes related to osteoblasts and osteoclasts22. Other important regulators of osteogenic functions. WWP1 studies have documented that the fine-tuning effects of accelerates the degradation of the osteogenic factor Runx2 miRNAs in bone remodeling and fracture repair are to repress its osteogenic function, and decreasing the achieved by controlling gene expression in bone cells23. WWP1 level in BMSCs may constitute a promising miR-19b is one of the components included in exosomes strategy to suppress the impairment of osteoblastic derived from MSCs24 and can promote fracture healing function27. The ubiquitin E3 ligase Smurf2 was also found during osteogenic differentiation25. Thus, we modulated to orchestrate osteoblast function by augmenting pro- miR-19b expression in BMSCs and in a mouse model to teasomal degradation of osteogenic proteins28. These explore the resulting effects on osteogenic factors, bone observations thus further corroborate our findings that cell mineralization and the healing status of modeled overexpression of WWP1 or Smurf2 led to degradation of fractures. The observations suggested that miR-19b their target protein KLF5 in BMSCs through ubiquitina- overexpression can facilitate the expression of osteo- tion to inhibit fracture healing. WWP1 and Smurf2 dis- genic factors (Runx2, ALP, and Collagen I), bone cell play similar inhibitory effects on KLF529. For example,

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 984

both use a binding domain and a catalytic mechanism to Conflict of interest enhance the degradation of KLF5, and the degradation of The authors declare no competing interests. KLF5 by both proteins is highly inhibited by the proteasome inhibitor MG132. The KLF5 protein is clearly Publisher’s note controlled by the ubiquitin-proteasome pathway, and Springer Nature remains neutral with regard to jurisdictional claims in WWP1 negatively regulates the function of KLF5 in gene published maps and institutional affiliations. regulation by controlling its ubiquitination and degrada- Supplementary information The online version contains supplementary 30 tion . Consistent with these observations, another study material available at https://doi.org/10.1038/s12276-021-00631-w. documented that Smurf2 specifically destabilizes KLF5, that the degradation of KLF5 by Smurf2 is disrupted by Received: 17 March 2020 Revised: 6 January 2021 Accepted: 29 January 2021. the proteasome inhibitor MG132, and that Smurf2 ubi- Published online: 25 May 2021 quitinates KLF511. Of note, KLF5 was reported to mediate the proliferative potential and osteogenic differentiation References of human periodontal ligament cells by restricting ALP 1. Einhorn, T. A. & Gerstenfeld, L. C. Fracture healing: mechanisms and inter- activity and Runx2 expression31. Our mechanistic studies ventions. Nat. Rev. Rheumatol. 11,45–54 (2015). 2. Garrison, K. R. et al. Bone morphogenetic protein (BMP) for fracture healing in further demonstrated that KLF5 knockdown delayed adults. Cochrane Database Syst. Rev. https://doi.org/10.1002/14651858. fracture healing by modulating the Wnt/β-catenin sig- CD006950.pub2CD006950 (2010). naling pathway. The activation of β-catenin can increase 3. Qiu, P. et al. Periosteal matrix-derived hydrogel promotes bone repair through an early immune regulation coupled with enhanced angio- and osteogenesis. bone formation and decrease bone resorption, resulting in Biomaterials 227, 119552 (2020). accelerated bone fracture healing, and thus, Wnt/ 4. Dulany, K., Hepburn, K., Goins, A. & Allen, J. B. In vitro and in vivo bio- β-catenin signaling activation should be sustained during compatibility assessment of free radical scavenging nanocomposite scaffolds 32 for bone tissue regeneration. J. Biomed. Mater. Res. A. 108, 301–315 (2020). fracture healing to intensify repair . In a previous mouse 5. Isaksson, H. et al. Remodeling of fracture callus in mice is consistent with model of closed tibial fracture, inactivation of β-catenin in mechanical loading and bone remodeling theory. J. Orthop. Res. 27, 664–672 chondrocytes impeded fracture healing due to restricted (2009). fi 33 6. Killock, D. Fracture repair: could blockade of Notch signalling promote fracture chondrogenesis and endochondral ossi cation . Our repair? Nat. Rev. Rheumatol. 10,3(2014). in vivo experiments in mouse models also substantiated 7. Toosi, S. & Behravan, J. Osteogenesis and bone remodeling: a focus on growth that BMSC-exos containing increased levels of miR-19b factors and bioactive peptides. Biofactors https://doi.org/10.1002/biof.1598 β (2019). enhanced fracture healing via the KLF5/ -catenin sig- 8. Liu, W. et al. Hypoxic mesenchymal stem cell-derived exosomes promote naling pathway by targeting WWP1 or Smurf2. bone fracture healing by the transfer of miR-126. Acta Biomater. 103, 196–212 Thus, we provided evidence confirming that upregula- (2020). β 9. Zhu, Y. et al. MiR-19b-3p regulates osteogenic differentiation of PDGFRalpha tion of miR-19b and activation of KLF5/ -catenin sig- (+) muscle cells by specifically targeting PTEN. Cell Biol. Int. 43,565–573 (2019). naling may potentially be clinically viable targets in the 10. Zhao,D.,Zhi,X.,Zhou,Z.&Chen,C.TAZantagonizestheWWP1-mediated treatment of fractures. More importantly, mesenchymal KLF5 degradation and promotes breast cell proliferation and tumorigenesis. Carcinogenesis 33,59–67 (2012). stem cell-derived exosomal miR-19b represses the 11. Du, J. X. et al. The E3 ubiquitin ligase, SMAD ubiquitination regulatory factor 2 expression of WWP1 or Smurf2 and elevates KLF5 (SMURF2), negatively regulates the PRO-proliferative KrüPpel-like factor 5 expression through the Wnt/β-catenin signaling pathway, (KLF5). Gastroenterology 140, 40354–40364 (2011). 12. Tang, J. et al. LncRNA PVT1 regulates triple-negative breast cancer through thereby facilitating fracture healing. As we further eluci- KLF5/beta-catenin signaling. Oncogene 37,4723–4734 (2018). date the epigenetic regulatory mechanisms underlying 13. Liu,Y.B.,Lin,L.P.,Zou,R.,Zhao,Q.H.&Lin,F.Q.Silencinglongnon-coding fracture healing, there is great potential for translational RNA MEG3 accelerates tibia fraction healing by regulating the Wnt/beta- catenin signalling pathway. J. Cell Mol. Med. 23,3855–3866 (2019). application of mesenchymal stem cell-derived exosomal 14. Mauney, J. R. et al. In vitro and in vivo evaluation of differentially demineralized miR-19b. cancellous bone scaffolds combined with human bone marrow stromal cells for tissue engineering. Biomaterials 26,3173–3185 (2005). Acknowledgements 15. Long, H. et al. miR-381 modulates human bone mesenchymal stromal cells We acknowledge and appreciate our colleagues for their valuable efforts and (BMSCs) osteogenesis via suppressing Wnt signaling pathway during atrophic comments on this paper. This study was supported by the Changsha Science nonunion development. Cell Death Dis. 10, 470 (2019). and Technology Project–Study on the Treatment of Traumatic Bone Defect 16. Zhu, L. P. et al. Hypoxia-elicited mesenchymal stem cell-derived exosomes with New Medical Magnesium Alloy Induced Bone Formation (No. kq1901050). facilitates cardiac repair through miR-125b-mediated prevention of cell death in myocardial infarction. Theranostics 8,6163–6177 (2018). 17. Kodama, A. et al. In vivo bioluminescence imaging of transplanted bone Author contributions marrow mesenchymal stromal cells using a magnetic delivery system in a rat Y.H. designed the study. Y.H., Y.Q.X., S.Y.F. and P.H. collated the data, carried out fracture model. J. Bone Jt. Surg. Br. 94,998–1006 (2012). data analyses and produced the initial draft of the manuscript. B.S. and J.D.N. 18. Maes, C. et al. Placental growth factor mediates mesenchymal cell develop- contributed to drafting the manuscript. All authors have read and approved ment, cartilage turnover, and bone remodeling during fracture repair. J. Clin. fi the nal submitted manuscript. Invest 116, 1230–1242 (2006). 19. Lacroix, D. & Prendergast, P. J. A mechano-regulation model for tissue dif- Data availability ferentiation during fracture healing: analysis of gap size and loading. J. Bio- The datasets generated and/or analyzed during the current study are available mech. 35,1163–1171 (2002). from the corresponding author on reasonable request.

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology Huang et al. Experimental & Molecular Medicine (2021) 53:973–985 985

20. Jin, Q. et al. USP7 cooperates with NOTCH1 to drive the oncogenic tran- 27. Shu,L.,Zhang,H.,Boyce,B.F.&Xing,L.UbiquitinE3ligaseWwp1negatively scriptional program in T-cell leukemia. Clin. Cancer Res. 25,222–239 (2019). regulates osteoblast function by inhibiting osteoblast differentiation and 21. Luo,Z.W.etal.Aptamer-functionalized exosomes from bone marrow stromal migration. J. Bone Min. Res. 28,1925–1935 (2013). cells target bone to promote bone regeneration. Nanoscale 11, 20884–20892 28. Liu, J. et al. Ubiquitin E3 ligase Itch negatively regulates osteoblast function by (2019). promoting proteasome degradation of osteogenic proteins. Bone Jt. Res. 6, 22. Ladang, A. et al. Evaluation of a panel of microRNAs that predicts fragility 154–161 (2017). fracture risk: a pilot study. Calcif. Tissue Int. 106,239–247 (2020). 29. Rotin,D.&Kumar,S.Physiologicalfunctions of the HECT family of ubiquitin 23. Zarecki, P., Hackl, M., Grillari, J., Debono, M. & Eastell, R. Serum microRNAs as ligases. Nat. Rev. Mol. Cell Biol. 10,398–409 (2009). novel biomarkers for osteoporotic vertebral fractures. Bone 130, 115105 30. Chen, C. et al. Human Kruppel-like factor 5 is a target of the E3 ubiquitin ligase (2020). WWP1 for proteolysis in epithelial cells. J. Biol. Chem. 280, 41553–41561 (2005). 24. Wang, X., Omar, O., Vazirisani, F., Thomsen, P. & Ekstrom, K. Mesenchymal stem 31. Gao,J.Y.,Yu,X.Q.&Wang,J.Q.[KLF5modulatesproliferationandosteogenic cell-derived exosomes have altered microRNA proﬁles and induce osteogenic differentiation of human periodontal ligament cells subjected to cyclic tensile differentiation depending on the stage of differentiation. PLoS ONE 13, stress]. Shanghai Kou Qiang Yi Xue 27,28–33 (2018). e0193059 (2018). 32. Bao, Q. et al. An appropriate Wnt/beta-catenin expression level during the 25. Furuta, T. et al. Mesenchymal stem cell-derived exosomes promote fracture remodeling phase is required for improved bone fracture healing in mice. Sci. healing in a mouse model. Stem Cells Transl. Med. 5,1620–1630 (2016). Rep. 7, 2695 (2017). 26. Lagos-Quintana,M.,Rauhut,R.,Lendeckel,W.&Tuschl,T.Identiﬁcation of 33. Huang, Y. et al. Inhibition of beta-catenin signaling in chondrocytes induces novel genes coding for small expressed RNAs. Science 294, 853–858 (2001). delayed fracture healing in mice. J. Orthop. Res. 30,304–310 (2012).

Ofﬁcial journal of the Korean Society for Biochemistry and Molecular Biology